ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time

MLLM
Re-assign community